If the fetch response is not a list but rather a list wrapped in an object, like

{
   data: [ blah, blah, blah ]
}

then it needs to be parsed before calling .length on it to decide whether there are more pages to load. This is exactly what Backbone's parse function is used for. The default implementation of parse is to just pass the response through, so there's no need to check if it's there before calling it.

I don't think what's sent to success and error should be changed because people probably already depend on it working a certain way. It's less likely that anyone depends on fetchSuccess's broken behavior in strict: false mode.

This fixes #7.